Influence of Impregnation Conditions on Tenoxicam Solubility and Loading into γ-Cyclodextrin Metal–Organic Frameworks: A Box–Behnken Design Approach

Abstract

Background/Objectives: γ-Cyclodextrin metal–organic frameworks (γ-CD-MOFs) are biocompatible porous crystalline materials that combine the advantages of both γ-cyclodextrins (γ-CDs) and MOFs, making them promising carriers for drug delivery. However, drug loading efficiencies into γ-CD-MOFs achieved by impregnation method involves complex interactions that necessitate further systematic exploration. This study aimed to determine the impregnation conditions that significantly impact tenoxicam (TNX) loading into γ-CD-MOFs and its aqueous solubility, and to identify the optimal possible conditions for maximizing both. Methods: A three-factor, three-level (3³) Box–Behnken factorial design technique was utilized. Results: Statistical analysis showed that TNX/γ-CD-MOF molar ratio exerted a significant positive effect on drug loading, whereas loading temperature and time have an insignificant effect. Additionally, while loading TNX into γ-CD-MOFs increased its water solubility, variations in the loading parameters did not produce a significant effect on this solubility. The impregnation conditions obtained from the numerical optimization step were a drug/MOF molar ratio of 1.99:1 at 29 ± 0.5 °C for 6 h, which experimentally showed TNX loading of 12.2 ± 1.55%. A discrepancy between the predicted and experimental drug-loading results was observed suggesting that the fitted model does not fully capture the complexity of the system, highlighting the need for experimental verification. Conclusions: This work delivers new insights into the impregnation factors governing TNX loading into γ-CD-MOFs and establishes a foundational framework for the future optimization of CD-MOFs-based drug formulations.

1. Introduction

Metal–organic frameworks (MOFs) are a fascinating class of hybrid materials that are composed of inorganic metal centers coordinated to organic linkers to form highly porous crystalline structures [1]. Depending on the choice of the metal ions and the organic linkers, MOFs may exhibit biocompatibility, biodegradability, high specific surface area, and tunable pore size, geometry, and functionality [2,3,4]. These unique chemical and physical properties make them ideal candidates for several potential uses such as biomedical applications and drug delivery [5,6]. γ-CD-MOFs are a green and biocompatible subclass of MOFs [7,8,9], which are easily synthesized by reacting γ-CD with biocompatible metal ions such as alkali metal cations (e.g., Na⁺ and K⁺) or transition metals (e.g., Fe⁺⁺) [10,11,12,13]. γ-CD-MOFs have a body centered cubic extended structure with two main kinds of cavities. The first void is composed of six γ-CD units coordinated by the metal ions to form large spherical pore with a dimension of 1.7 nm and aperture windows dimensions of 0.8 nm. The second pore is formed by the face-to-face γ-CD pairs with a 0.8 nm diameter (Figure 1) [7]. The structure features an interconnected network of channels, enabling the large spherical pores to act as confined nanoreactors [14].

γ-CD-MOFs combine the intrinsic properties of γ-CDs (e.g., inclusion capabilities) with the structural characteristics of MOFs. Therefore, they are expected to provide better encapsulation and improved solubility and bioavailability of different drugs such as TNX among other non-steroidal anti-inflammatory drugs (NSAID) [10,15].

TNX is a lipophilic NSAID with anti-inflammatory, analgesic, and antipyretic properties that is often used in the treatment of rheumatic and arthritic diseases [16]. According to the biopharmaceutical classification system (BCS), TNX is classified as Class II drug where it has dissolution-limited bioavailability due to its poor water solubility [17]. Additionally, because it is a weakly acidic drug (pK_a = 5.3), it has a limited number of counter-ions with which it can form stable, pharmaceutically acceptable salts [18]. TNX has a molar mass of 337.37 g.mol⁻¹ and molecular dimensions of approximately 1.3 nm in length and 0.5 nm in width (Figure 2), measured using Mercury software (version 3.8, CCDC, Cambridge, UK) [19].

Drug incorporation into γ-CD-MOFs can be achieved using different methods including, but not limited to, co-crystallization during MOFs synthesis and impregnation of pre-synthesized γ-CD-MOF crystals [15,20,21,22]. However, numerous studies have shown that impregnation-based loading of many drugs into γ-CD-MOFs is generally low [21]. For example, Liu et al. reported that the loading of the NSAIDs piroxicam, ketoprofen, and meloxicam were 8.44, 7.40, and 3.22%, respectively [23]. They also reported that the loading of other drugs such as paracetamol, metronidazole and caffeine were only 0.3, 0.02, and 0.01%, respectively [23]. Additionally, it was reported that drug encapsulation into γ-CD-MOFs is strongly dependent on loading conditions, which can significantly influence the frameworks entrapment efficiency [15,24]. For instance, Li et al. demonstrated that altering the impregnation solvent caused substantial changes in the ibuprofen loading capacity of γ-CD-MOFs [24]. This in fact highlights the need for judicious selection of the appropriate impregnation conditions and necessitates further investigations.

The present study aimed to verify the impregnation conditions that significantly influence TNX loading into γ-CD-MOFs and its water solubility, using a 3³ Box–Behnken factorial design. The factors investigated were TNX/γ-CD-MOF molar ratio, loading temperature, and loading time. Additionally, the study sought to identify the optimal conditions for achieving maximum TNX loading and solubility. Furthermore, the effect of crystal size on drug loading and dissolution was also examined.

2. Materials and Methods

2.1. Materials

γ-CD was purchased from Shandong Zhi Shang Chemical Co., LTD (Jinan, China). TNX was kindly provided by Egyptian International Pharmaceutical Industries Co., EIPICo (Cairo, Egypt). Potassium hydroxide and cetyltrimethylammonium bromide (CTAB) were purchased from Sigma-Aldrich Chemie, GmbH (Steinheim, Germany) and used without future purification. All other solvents and chemicals were of analytical grade.

2.2. Synthesis of γ-CD-MOFs

γ-CD-MOF crystals were synthesized using a modified vapor diffusion method (Figure 3) [23]. Briefly, in 5 mL of distilled deionized water, γ-CD (162 mg, 0.125 mmol) was reacted with KOH (56 mg, 1 mmol) at a ratio of 1:8. Any possible dust or impurities were removed by filtering this solution through a 0.45 µm syringe filter into a beaker containing 0.5 mL of methanol (MeOH) [25]. The beaker was placed in a MeOH-filled (20 mL) glass vessel, which was sealed, placed in an oven at 50 °C, and MeOH vapors were allowed to diffuse into the reaction mixture for 6 h. The formed large crystals (L-γ-CD-MOF) were collected by centrifugation (2000 rpm for 2 min at 25 °C), washed three times and activated by solvent-exchange using MeOH. The supernatant was collected, CTAB was added in a concentration of 8 mg/mL, and the mixture was incubated overnight at room temperature (RT). The addition of CTAB as a modulator was to slow crystal growth, trigger fast precipitation, consequently, obtain monodispersed smaller crystals (S-γ-CD-MOF) [26,27]. The formed small γ-CD-MOF crystals precipitate was centrifuged and washed three times then activated with MeOH. It is worth noting that the activation step involves using a solvent with a lower boiling point to unblock the pores by removing any unreacted materials or residual solvents. Therefore, all formed crystals were submerged in MeOH for 24 h, during which the solvent was regularly decanted and replaced frequently, followed by drying under vacuum at 40 °C overnight. The activated dried crystals were stored in a tightly closed container at 4 °C until further use.

To calculate the % yield (Y) for both L-γ-CD-MOF and S-γ-CD-MOF crystals, the ratio between the wight of dried crystals, and that of starting materials (metal salt and γ-CD) was calculated using Equation (1) [10].

$$\mathrm{Y}\left(\%\right)=\left({\displaystyle \frac{\mathrm{Weight}\mathrm{of}\mathsf{\gamma }\text{-}\mathrm{CD}\text{-}\mathrm{MOF}\mathrm{prepared}}{\mathrm{Total}\mathrm{weight}\mathrm{of}\mathrm{reagents}}}\right)\times 100$$

(1)

2.3. Factors Governing TNX Loading into S-γ-CD-MOF

In this study, TNX loaded crystals (TNX-γ-CD-MOFs) were obtained using the impregnation technique. Hence, to identify the impregnation conditions that significantly affect drug loading into S-γ-CD-MOF and its water solubility, 3³ Box–Behnken factorial design was used utilizing a statistical package (Statgraphics Plus, version 5). TNX/γ-CD-MOF molar ratio (X1), loading temperature (X2) and time (X3) were considered the independent factors and were examined at three levels (

Table 1. Variables in 3³ Box–Behnken factorial design of TNX loading into γ-CD-MOFs.

Independent Variable: Factor	Low (−1)	Middle (0)	High (1)
X1: TNX/S-γ-CD-MOF (molar ratio)	1	1.5	2
X2: loading temperature (°C)	25	37.5	50
X3: loading time (h)	2	4	6
Dependent variable: Response
Y1: drug loading (%)
Y2: TNX solubility in water (µg/mL)

).

The experimental design suggested a total of 13 experiments, which are shown in

Table 2. Impregnation conditions for TNX loading into S-γ-CD-MOF.

Formula	TNX/MOF (X1, Molar Ratio)	Temperature (X2, °C)	Time (X3, h)
F1	1:1	25	4
F2	1.5:1	25	2
F3	1.5:1	25	6
F4	2:1	25	4
F5	1:1	37.5	2
F6	1:1	37.5	6
F7	1.5:1	37.5	4
F8	2:1	37.5	2
F9	2:1	37.5	6
F10	1:1	50	4
F11	1.5:1	50	2
F12	1.5:1	50	6
F13	2:1	50	4

and were conducted in duplicates (Formulations F1 to F13). Statistical models including main, quadratic, and interactive modes were studied to evaluate the effect of the independent factors on % payload (%; Y1) and TNX aqueous solubility after inclusion in the framework (µg/mL; Y2).

The impregnation approach was performed experimentally where the calculated amount of TNX that satisfies each specified TNX/MOF molar ratio was weighed, added to 10 mL of MeOH, followed by the addition of 100 mg of activated S-γ-CD-MOF [23]. The suspension was shaken in a water bath at 100 rpm for 2, 4, and 6 h at 25, 37.5, and 50 ± 0.5 °C. The drug-loaded crystals (TNX-S-γ-CD-MOF) were harvested by centrifugation, washed four times with an equivalent volume of the incubation solvent, and dried overnight at 40 °C. The washing step is to ensure removing any surface adsorbed drug molecules prior to further processing or analysis [11,24]. The amount of loaded drug was quantified directly by dissolving the loaded crystals in distilled water with the aid of sonication, filtering through 0.45 µm syringe filters, diluting, and measuring absorbance by UV-visible spectrophotometer (Libra S22, Biochrom, Cambridge, England) at ʎ_max 370 nm. A standard calibration curve was constructed over a concentration range of 4 to 20 µg/mL. The percent payload was calculated using Equation (2) [28].

$$\mathrm{Payload}\left(\%\right)=\left({\displaystyle \frac{\mathrm{Encapsulated}\mathrm{drug}\left(\mathrm{mg}\right)}{\mathrm{Drug}\mathrm{loaded}\mathrm{MOF}\left(\mathrm{mg}\right)}}\right)\times 100$$

(2)

2.4. Effect of Impregnation Conditions on TNX Aqueous Solubility After Inclusion

To study the effect of encapsulating TNX into S-γ-CD-MOF using different impregnation conditions on its water solubility, the shake-flask method was utilized with some modifications [14]. Here, TNX-S-γ-CD-MOF crystals loaded using different impregnation conditions and containing equivalent amount of loaded drug were weighed and placed in stoppered 25-mL conical flasks containing 10 mL of distilled water. The flasks were shaken in a thermostated horizontal shaker (SW22, Julabo GmbH, Seelbach, Germany) at 150 rpm and 25 ± 0.5 °C for 72 h. After reaching equilibrium, the different solutions were centrifuged, and aliquots of the supernatants were diluted to concentrations appropriate for UV-spectrophotometric determination at ʎ_max 370 nm (after filtration through 0.45 µm syringe filter). The reported values are the average of duplicate experiments.

2.5. Toward Optimizing TNX Impregnation Conditions

In a trial to optimize TNX loading into S-γ-CD-MOF crystals as well as its water solubility, a combination of the independent variables satisfying the desirability criteria was explored using multiple-response analysis based on maximum drug loading and water solubility. It is worth noting that TNX was loaded into both S-γ-CD-MOF and L-γ-CD-MOF using the impregnation conditions obtained from this numerical optimization step. Here, “numerical optimization” refers to the desirability-based optimization function of the response surface model.

2.6. Characterization of γ-CD-MOFs and TNX-γ-CD-MOFs

2.6.1. Powder X-Ray Diffraction (PXRD)

PXRD data were collected using Rigaku miniflex 300/600 (Tokyo, Japan), equipped with a Cu Kα radiation source (λ = 1.540562 Å) at an excitation voltage of 40 kV and a current of 15 mA. The samples were fixed on glass holders and scanned at a rate of 5°/min between 2θ ranging from 5° to 30°. The experimental acquired powder patterns were used to verify crystallinity, purity and stability of γ-CD-MOFs and TNX-γ-CD-MOFs by comparing them to the calculated patterns [29] that were acquired using Mercury software [19].

2.6.2. Morphological Characterization

Morphological characteristics of the as-synthesized and drug-loaded large and small crystals were studied using scanning electron microscopy (SEM) (JSM-IT500HR SEM, Jeol Inc., Peabody, MA, USA). The samples were collected on carbon tape and coated with a 2 nm layer of platinum using a JEC-3000FC auto fine coater (JEOL Inc., Peabody, MA, USA) to enhance conductivity and imaging quality. The samples were then imaged with the SEM, with measurements taken at an accelerating voltage of 5 kV.

2.6.3. Thermal Analysis

Thermogravimetric analysis (TGA) was performed using a TGA/DSC1 analyzer (Mettler Toledo, Greifensee, Switzerland). Here, pure TNX, large and small as-synthesized and drug- loaded γ-CD-MOF crystals were analyzed under the same conditions and heated from RT to 400 °C at a heating rate of 10 °C/min under constant N₂ flow. TGA was used to determine moisture content, thermal stability, and decomposition temperature of the crystals.

2.6.4. Fourier-Transform Infrared Spectroscopy (FTIR)

To confirm the chemical structure of TNX, γ-CD, blank, and drug-loaded γ-CD-MOFs, Thermo smart ATR IS20 Spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) was used for the analysis of the FTIR. The spectral resolution was configured at 4 cm⁻¹, and 32 scans were conducted for each sample, which was analyzed across a wavenumber range of 4000 to 650 cm⁻¹. Here, samples were placed on the sampling area, one sample at a time in small quantities (~7 mg).

2.7. In Vitro Dissolution Studies

TNX in vitro dissolution studies from large and small crystals loaded using the impregnation conditions obtained from the numerical optimization step were performed using USP dissolution apparatus II (paddle method) on Pharma Test dissolution tester (DT 70 drive unit; Pharma Test Apparatebau AG, Hainburg, Germany). Five hundred mL of phosphate-buffered saline (PBS), pH 6.8 were used in each flask, the temperature was maintained at 37 ± 0.5 °C, and the paddle speed was set at 100 rpm. Twenty milligrams of TNX, or an equivalent weight of loaded crystals, were dispersed over the dissolution medium. At pre-set time intervals (5, 10, 15, 30, 45, and 60 min), a 2 mL-sample was withdrawn from each flask and replaced with an equivalent amount of prewarmed PBS. The dissolution media were kept under sink conditions, and the study was performed in triplicate. The withdrawn samples were filtered using 0.45 um syringe filters and analyzed by UV-visible spectrophotometer (Libra S22, Biochrom, Cambridge, England) at λ_max 370 nm where it was verified that there was no interference from other ingredients with drug assay. A standard calibration curve was constructed over a concentration range of 4 to 20 µg/mL. The dissolution results were analyzed and expressed as mean and standard deviation (SD), and the percentage of dissolved TNX was determined as a function of time.

2.8. Statistical Analysis

Statistical analysis was studied using one-way analysis of variance (ANOVA) by means of Statgraph Centurion Statistical Software (version 17, Statgraph Inc., Houston, TX, USA) with the Tukey–Kramer multiple assessments. Additionally, Student’s t-test was used for analysis. All values were calculated as their mean ± SD, and all statistically significant differences were anticipated when p < 0.05.

3. Results and Discussion

The aim of the current study was to identify the impregnation conditions that significantly affect both the loading of TNX (a poorly water-soluble NSAID) into γ-CD-MOFs and its aqueous solubility after inclusion into the framework. Consequently, a 3³ Box–Behnken factorial design was used utilizing a statistical package (Statgraphics Plus, version 5) to evaluate the impact of TNX/γ-CD-MOF molar ratio (X1), loading temperature (X2) and time (X3) on the percent drug payload (Y1) and its water solubility (Y2). It is worth mentioning that the present work intentionally focused on a limited, but dominant, set of factors that are widely reported in drug loading studies to reduce model complexity and establish a robust baseline understanding of TNX loading into γ-CD-MOFs. The selected temperatures are primarily expected to influence both drug solubility and molecular diffusion or mobility within the pores while preserving the structural integrity of the framework [30]. On the other hand, impregnation time influences host–guest interaction equilibrium and determines the extent to which equilibrium loading is achieved [31]. Additionally, MeOH was selected in this work as the encapsulation solvent since previous studies have demonstrated that alcohols, particularly MeOH and ethanol, provide superior encapsulation efficiency in CD-based MOFs [32]. Notably, MeOH has been reported to consistently yield higher loading capacity by promoting stronger host–guest interactions and minimizing competitive solvent inclusion [33].

Accordingly, γ-CD-MOF crystals were first synthesized using a modified vapor diffusion method. The yield obtained was around 63.33 ± 10.4% for L-γ-CD-MOF and 22.4 ± 4.6% for S-γ-CD-MOF (n = 3), values that indicate the efficiency of the adapted preparation method [10].

3.1. Factors Governing TNX Loading into S-γ-CD-MOF

Here, only small γ-CD-MOF crystals were used in the experimental design to ensure homogeneous drug loading, faster equilibration, reduced diffusion heterogeneity, and robust response surface modeling [34]. Consequently, the effect of different impregnation conditions on the % payload of TNX into S-γ-CD-MOF and its solubility after inclusion is shown in

Table 3. TNX % payload in S-γ-CD-MOF and its water solubility after using different TNX/γ-CD-MOF molar ratios, loading temperatures, and time as the impregnation conditions (n = 2).

Formula	Drug Payload (Y1, %)	TNX Solubility (Y2, µg/mL)
F1	6.35 ± 0.71	56.46 ± 4.97
F2	16.19 ± 1.70	34.46 ± 3.13
F3	14.22 ± 1.93	46.50 ± 4.89
F4	20.02 ± 1.86	54.81 ± 2.63
F5	6.84 ± 0.69	49.34 ± 3.20
F6	7.74 ± 0.43	46.94 ± 4.27
F7	11.98 ± 1.40	48.80 ± 4.74
F8	26.75 ± 0.39	51.20 ± 2.40
F9	24.29 ± 1.96	54.98 ± 4.97
F10	5.74 ± 0.46	53.39 ± 0
F11	9.68 ± 0.78	37.31 ± 1.11
F12	9.68 ± 0.47	47.92 ± 2.06
F13	18.11 ± 1.86	34.03 ± 4.27

. The obtained payload values ranged from 5.74 ± 0.46 to 26.75 ± 0.39 percent for F10 and F8, respectively. Across the different impregnation conditions, it is clearly evident that drug loading efficiency is mainly governed by the drug-to-carrier ratio.

Statistical analysis revealed that drug-to-carrier ratio showed a significant positive effect on drug loading percentage (p = 0.0062), while the loading temperature and time exhibited insignificant antagonistic effects (p > 0.05). This could be observed in the standardized Pareto charts for the effects of the independent variables on drug loading efficiency represented as the % payload (Figure 4).

Enhanced drug loading was observed with increasing TNX/MOF ratio to 2:1, especially for formulations F8 and F9 (26.75 ± 0.39 and 24.29 ± 1.96, respectively) that were performed at a loading temperature of 37.5 ± 0.5 °C regardless of the loading time.

3.2. Effect of Impregnation Conditions on TNX Aqueous Solubility After Inclusion

TNX belongs to Class II according to the BCS, thus it is a poorly water-soluble drug with high permeability [17]. TNX exhibited a water solubility of only 15.64 ± 0.0 µg/mL, a value consistent with previously reported data [35]. A noticeable increase in this value was observed after the drug inclusion into the framework regardless of the impregnation conditions as shown in Table 3. The solubility measurements varied from 34.03 ± 4.27 to 56.46 ± 4.97 µg/mL for F13 and F1, respectively. By applying the 3³ Box–Behnken factorial design it was found that the independent loading parameters generally did not exhibit significant actions on drug solubility (p-values > 0.05), which might be attributed to the opposite actions of positive effects of impregnation time and quadratic effect of TNX/MOF ratio against the antagonistic effects of impregnation temperature, its quadratic effect, TNX/MOF ratio and impregnation time quadratic effect. This could be observed in the standardized Pareto charts for the effects of the independent variables on TNX water solubility (Figure 5).

It is worth mentioning that the enhanced water solubility of TNX after loading into the framework, regardless of the impregnation conditions, can be attributed to the initial inclusion complexes formation with the CD units in the framework [16]. Additionally, TNX nanoclusters form within the framework’s nanoreactor cavities upon loading, followed by their release due to the hydrolytic instability and water solubility of the framework [14,26]. Thus, it is suggested that CD-MOFs-mediated solubilization is primarily driven by host–guest interactions rather than by impregnation factors variations.

3.3. Toward Optimizing TNX Impregnation Conditions

The factorial design technique was proven to be effective in pharmacy practice, in general, especially in formulations optimization [36]. To the best of our knowledge, this study represents the first attempt to apply a Box–Behnken factorial design to systematically explore and optimize impregnation conditions for drug loading into γ-CD-MOFs. Herin, this technique was utilized toward optimizing the impregnation conditions for achieving the highest TNX payload and its maximum water solubility. In addition to predicting the expected response values under these conditions, followed by experimental validation. Therefore, the impregnation conditions obtained from the numerical optimization step suggested the use of TNX/γ-CD-MOF molar ratio of 1.99:1 at 29 °C for 6 h. These conditions were studied as a part of the experimental design, and the results were verified by comparing the program predicted values and the observed experimental ones (

Table 4. The suggested impregnation conditions, the predicted response values, and the experimentally observed values of TNX % payload and solubility.

Factors (X)	Suggested Value	Response	Desirability	Predicted	Observed
TNX/S-γ-CD-MOF (X1, Molar ratio)	1.99
Loading temperature (X2, °C)	29	Payload (%)	Maximize	22.895	12.2 ± 1.55
Loading time (X3, h)	6	Solubility (µg/mL)	Maximize	56.46	55.54 ± 3.2

).

The % payload in S-γ-CD-MOF was found to be 12.2 ± 1.55%, which differs from the predicted value of 22.895%. Additionally, TNX aqueous solubility was found to be 55.54 ± 3.2 µg/mL, which is close to the predicted value of 56.46 µg/mL. While the predicted solubility showed good agreement with the experimental results, the notable deviation observed for TNX payload highlights inherent limitations of Box–Behnken design-based optimization models and warrants further discussion.

The fit model suggested by the software was linear with high R² value of 0.961, adjusted R² of 0.8444, and predicted R² of 0.737. The difference between adjusted and predicted R² is less than 0.2, which is reasonable. The impact of the interactive and quadratic models on the tested responses and their effect on the dependent parameters was studied and the results for TNX payload are listed in the ANOVA table (

Table 5. ANOVA table for the effects of independent variables, their interactive, and quadratic models on TNX payload.

Source	Sum of Squares	Degree of Freedom (df)	Mean Square	F-Value	p-Value
Model	560.81	9	62.31	8.24	0.0548	Not significant
A-Drug/MOF ratio	488.28	1	488.28	64.53	0.0040
B-Loading Temperature	23.01	1	23.01	3.04	0.1796
C-Loading time	1.56	1	1.56	0.2057	0.6810
AB	0.4309	1	0.4309	0.0570	0.8268
AC	2.83	1	2.83	0.3740	0.5841
BC	0.9696	1	0.9696	0.1281	0.7441
A2	11.75	1	11.75	1.55	0.3012
B2	6.55	1	6.55	0.8652	0.4209
C2	10.64	1	10.64	1.41	0.3211
Residual	22.70	3	7.57
Cor Total	583.51	12

).

The sum squared residual is 22, indicating good model fit, which is proved by the high R² value. Additionally, adequate precision is important in Design of Experiments (DoE) and it measures the signal to noise ratio to confirm if a statistical model can reliably predict responses. It ensures the model’s “signal” (range of predictions) is much stronger than its “noise” (prediction error), with a ratio > 4 often indicating good predictive power for optimization. For this design, the fit statistics analysis for the quadratic model showed that the adequate precision equals to 9.17, indicating an adequate signal, and the model could be used to navigate the design space. The adequacy of the response surface model for TNX payload was evaluated using standard regression fit statistics, as summarized in

Table 6. Fit statistics of the response surface model for TNX payload (Y1).

R2 (Coefficient of Correlation)	Adjusted R2	Predicted R2	Adequate Precision	Standard Deviation (Std. Dev.)	Mean	Coefficient of Variation (C.V. %)
0.9611	0.8444	NA ⁽1⁾	9.1706	2.75	13.66	20.14

⁽¹⁾ Not Available (missing value).

.

The statistical equation for payload is described in Equation (3) where (A) is drug/MOF ratio, (B) is the loading temperature and (C) is the loading time.

Payload = 11.98 + 7.81A − 1.7B − 0.4411C − 0.33AB − 0.84AC + 0.49BC + 2.27A2 − 1.69B2 + 2.16C2

(3)

Drug loading into γ-CD-MOFs via impregnation technique is inherently complex, and influenced by multiple physicochemical processes [37,38], whereas Box–Behnken optimization model uses mathematical equations that may not fully represent the real nonlinear behavior involved in this loading process. The complexity of the process could be attributed to several factors that are related to the nature of γ-CD-MOFs, its interaction with the solvent and/or the guest molecules [28]. γ-CD-MOFs may exhibit slight variation in its crystal size and/or surface area even when the same synthesis conditions are used. This could be attributed to some defects such as missing linkers or metal nodes that lead to uncoordinated sites, pore collapse, or micro-cracks [39,40]. Moreover, γ-CD-MOF is hydrolytically unstable, which may further contribute to this variation in crystallinity and framework swelling or partial collapse [26]. In addition, pores accessibility of the material may differ due to factors related to the solvent including solvent penetration behavior, and its competition with the drug molecules to occupy the pores [41]. Furthermore, drug diffusion into the framework could be slow [42], the adoption process could be partially reversible, and some pores may not be accessible to the drug [43]. Additionally, other uncontrolled factors or interactions that are not included in the design such as humidity, solvent evaporation, and experimental variability during drug loading may also influence the response and contribute to preventing the mathematical prediction from fully capturing the real behavior of the framework.

In summary, the results indicate that the Box–Behnken design successfully identified the factors influencing TNX loading into γ-CD-MOFs. However, the observed discrepancy between experimental and predicted payload values does not invalidate the model. It rather indicates the need for additional confirmatory experimental runs to further refine and validate the predicted conditions or to include more factors in the study.

In contrast, the quadratic model was employed to examine the extent to which the independent factors and their interactivity influence drug solubility. Unlike the payload response, the model exhibited a poor fit with a low R² of 0.607 and a high sum squared residual of 274.85. Moreover, the adequate precision equals 2.22, which is less than 4, indicating inadequate signal and inability to use both linear and quadratic models to navigate the design space. The impact of the interactive and quadratic models on the tested responses and their effect on the dependent parameters was studied and the results for TNX solubility are listed in the ANOVA table (

Table 7. ANOVA table for the effects of independent variables, their interactive and quadratic models on TNX solubility.

Source	Sum of Squares	Degree of Freedom (df)	Mean Square	F-Value	p-Value
Model	424.97	9	47.22	0.5154	0.8063	Not significant
A-Drug/MOF ratio	15.42	1	15.42	0.1683	0.7092
B-Loading Temperature	47.94	1	47.94	0.5233	0.5217
C-Loading time	72.09	1	72.09	0.7869	0.4404
AB	78.54	1	78.54	0.8572	0.4228
AC	9.55	1	9.55	0.1043	0.7680
BC	0.5057	1	0.5057	0.0055	0.9454
A2	56.49	1	56.49	0.6166	0.4896
B2	38.35	1	38.35	0.4186	0.5637
C2	22.71	1	22.71	0.2479	0.6528
Residual	274.85	3	91.62
Cor Total	699.82	12

).

The adequacy of the response surface model for TNX solubility was evaluated using standard regression fit statistics, as summarized in

Table 8. Fit statistics of the response surface model for TNX solubility (Y2).

R2 (Coefficient of Correlation)	Adjusted R2	Predicted R2	Adequate Precision	Standard Deviation (Std. Dev.)	Mean	Coefficient of Variation (C.V. %)
0.6073	−0.5710	NA ⁽1⁾	2.2286	9.57	47.40	20.20

⁽¹⁾ Not Available (missing value).

.

The one-factor plot shows that the drug/MOF ratio (A), as well as the loading temperature (B) have negative effects on drug solubility. While the loading time (C) has a positive effect (Figure S1). This also can be determined from Equation (4).

Solubility = 48.8 − 1.39A − 2.45B + 3C − 4.43AB + 1.55AC − 0.355BC + 4.97A2 − 4.1B2 − 3.15C2

(4)

Likewise, the impregnation conditions obtained from the numerical optimization step were also applied to load the drug into L-γ-CD-MOF where the % payload was found to be 13.5 ± 0.5%. Statistical analysis showed that, according to conventional criteria, the difference % payload between small and large crystals is not statistically significant, with a two-tailed p-value equals to 0.2813. It is worth noting that the accessible pore volume and host–guest interactions governing TNX incorporation are comparable in both small and large γ-CD-MOF crystals due to their identical internal framework structure and chemical composition [34]. They differ mainly in their physical characteristics, such as particle size and surface area. Subsequently, host–guest interactions and the accessible pore volume governing TNX incorporation are comparable in both systems. This resulted in similar drug loading capacities under equilibrium impregnation conditions with no statistically significant difference in the % payload.

3.4. Characterization of γ-CD-MOFs and TNX-γ-CD-MOFs

3.4.1. Powder X-Ray Diffraction (PXRD)

Figure 6 shows the PXRD diffractograms of TNX, γ-CD-MOFs, and TNX-γ-CD-MOFs loaded using the impregnation conditions obtained from the numerical optimization step. The agreement between the calculated PXRD patterns of γ-CD-MOFs [29] and those of the free and drug-loaded crystals verifies their crystallinity, purity of the as-synthesized crystals, and the stability of the samples following drug inclusion [15]. The slight weakening of diffraction peaks after loading is likely due to partial γ-CD unit leakage or loss during the impregnation procedure [14]. Finally, the PXRD pattern of TNX showed its characteristics intrinsic peaks. These peaks disappeared in TNX-γ-CD-MOFs patterns since the drug is loaded inside the cages of the framework.

3.4.2. Morphological Characterization

SEM was used to study the morphological characterization of the as-synthesized and drug-loaded crystals. As seen in Figure 7a,c, both as-synthesized small and large γ-CD-MOFs crystals showed the typical cubical structure with smooth surfaces, which is consistent with the literature [22]. When compared to large crystals, the small ones synthesized by adding CTAB as a modulator had a more uniform particle size and shape (Figure 7a). Additionally, some large crystals were fused together and few displayed morphological deformations (Figure 7c).

After TNX loading, MOF crystals exhibited a visible color change from transparent to yellow, corresponding to the color of the drug. Furthermore, several TNX-S-γ-CD-MOF crystals exhibited minor surface cracks (Figure 7b), while some TNX-L-γ-CD-MOF crystals appeared fractured (Figure 7d). Generally, both small and large crystals maintained their cubical shape implementing that TNX loading did not significantly influence crystals morphology.

Although particle size was not quantified in this study, SEM images qualitatively indicate a clear difference in crystals size between S-γ-CD-MOF and L-γ-CD-MOF, consistent with the synthesis conditions reported in the literature [33].

3.4.3. Thermal Analysis

TGA was used to assess moisture content and thermostability of the as-synthesized and drug-loaded large and small crystals by heating from RT to 400 °C at 10 °C/min under nitrogen flow (Figure 8).

Here, the results demonstrate that both activated large and small crystals of γ-CD-MOFs undergo weight loss within the range of 30–100 °C with an initial weight loss of around 13 wt%, presumably attributed to the residual solvent and moisture in the framework that evaporates upon heating [44]. However, the entrapped solvent experienced partial replacement by TNX molecules during impregnation and the solvent contents were around 5 wt% and 4 wt% in TNX-L-γ-CD-MOF and TNX-S-γ-CD-MOF, respectively. As far as thermal stability is concerned, both as-synthesized and drug-loaded large and small crystals exhibit robust thermal stability, decomposing at around 257 °C. The weight loss initiated under this temperature could largely be attributed to the degradation of loaded drug molecules and the collapse of the three-dimensional framework due to the cleavage of metal–organic bonds and ligand decomposition [45].

It is worth noting that the decomposition point of free TNX was observed at 132 °C. At this temperature, the loaded frameworks showed a slower weight loss, until reaching the framework decomposition temperature. This could be an indication that the CD-MOFs influenced the decomposition kinetics and improved the thermal stability of the encapsulated drug where TNX decomposition was shifted to a higher temperature [46,47].

3.4.4. Fourier-Transform Infrared Spectroscopy (FTIR)

As shown in Figure 9, the FTIR was used to confirm the chemical structure of different studied materials. γ-CD-MOFs FTIR spectra showed that it kept its similarity to its building units (γ-CD). Additionally, drug-loaded MOFs spectra were completely consistent with that of blank MOFs, indicating that the loading process caused no disruption to the crystal structure. Furthermore, at 3000–3600 cm⁻¹, a characteristic peak was observed, which corresponds to the glucose unit -OH stretching vibration [22]. The -CH₂ stretching vibration was seen as a peak at 2928 cm⁻¹, and the C-O-C and C-O stretching vibrations were recorded at 945–1070 cm⁻¹. Importantly, no new absorption bands appear, suggesting that TNX is incorporated through non-covalent host–guest interactions rather than coordination or covalent bonding with the frameworks. However, the broad –OH stretching band becomes wider and slightly shifted, further supporting the formation of intermolecular hydrogen bonds between TNX and the γ-CD-MOFs (host frameworks).

Furthermore, the drug-loaded CD-MOFs FTIR spectra did not show the typical characteristic peaks of TNX [48], such as the S=O stretching vibration at 1150 cm⁻¹ and C-N stretching vibration at 1560 cm⁻¹ (Figure 2), which could be attributed to the low drug loading values of TNX of about 12.2 ± 1.55% and 13.5 ± 0.5% in small and large crystals, respectively. Overall, the FTIR results confirm that TNX is successfully incorporated into the CD-MOFs matrix through hydrogen bonding and inclusion within cyclodextrin cavities, while preserving both the chemical integrity of the drug and the structural frameworks of the MOFs.

3.5. In Vitro Dissolution Studies

Given the poor water solubility of TNX, which is the main cause of its limited bioavailability, it was essential to evaluate the in vitro dissolution profile of the drug from crystals loaded using the impregnation conditions obtained from the numerical optimization step. Furthermore, assessing the influence of crystal size on TNX dissolution behavior was also performed. Therefore, the dissolution rate of pure TNX and the in vitro drug dissolution from TNX-L-γ-CD-MOF and TNX-S-γ-CD-MOF loaded crystals were studied and are shown in Figure 10. Initially, TNX standard calibration curve was constructed, and it was linear over a concentration range of 4 to 20 µg/mL, with a determination coefficient 99.97 [36]. Moreover, the dissolution rate of pure TNX remained limited due to its poor water solubility, with a dissolution of less than 75% within 60 min, a finding that aligns with values documented in the literature [49]. Conversely, TNX was rapidly dissolved from the small crystals, reaching nearly 100% within the first 5 min. Also, the drug exhibited rapid dissolution from large crystals, in which 92.4% of the loaded drug dissolved within 60 min. Moreover, TNX dissolution rate from small CD-MOF crystals was significantly higher in comparison to its dissolution from the large ones (Test Statistic F was 6.14, and p-value was 0.048). This statistical comparison was inferred from the slope of the dissolution profiles and refers to differences in the dissolution rate. It is worth noting that the stated dissolution percentage represents the amount of drug dissolved at a specific time point, while the dissolution rate refers to the slope of the dissolution profile over time.

The rapid dissolution observed for the small crystals is likely due to their larger effective surface area compared with the large crystals. Additionally, the enhanced drug dissolution from both large and small crystals can be attributed to the hydrolytic instability and high water solubility of the γ-CD-MOFs [26,27]. Thus, upon contact with aqueous dissolution media, total frameworks disassembly occurs, which facilitates the liberation of TNX nanoclusters formed within the framework’s nanoreactor cavities [14].

In this study, γ-CD-MOFs–based drug delivery system hydrolytic instability is a functionally relevant feature that directly contributes to the rapid dissolution of the drug. On the other hand, for practical applications, the hydrolytic instability of the frameworks may be considered as a limitation, especially in the biomedical field for long-term stability and safety [27]. Therefore, different trials were reported in the literature to improve the framework water stability by cross-linking or surface modification but still benefit from the intrinsic properties of CD in forming inclusion complexes and enhancing water solubility of poorly water soluble drugs [15,26,50,51].

Taken together, these findings demonstrate that γ-CD-MOFs encapsulation significantly enhances TNX dissolution, with crystals size playing a major role in modulating its dissolution profile, which could be beneficial in improving the bioavailability of the drug [22].

4. Conclusions

The aim of the current study was to identify the impregnation conditions that significantly impact TNX loading into γ-CD-MOFs and its water solubility using a 3³ Box–Behnken factorial design utilizing a statistical package (Statgraphics Plus, version 5). The impregnation conditions under investigation were TNX/MOF molar ratio, loading solution temperature and loading time. The statistical analysis of the Box–Behnken design indicated that TNX/MOF molar ratio has a significant effect on the % drug payload, while loading temperature and time has an insignificant effect. Additionally, it was found that the three impregnation conditions have an insignificant effect on drug solubility after inclusion. This study highlights the importance of meticulously selecting the impregnation conditions for γ-CD-MOFs. However, when the impregnation conditions obtained from the numerical optimization step (TNX/MOF molar ratio of 1.99:1 at 29 °C for 6 h) were experimentally tested, the measured drug loading differed from the predicted value. This discrepancy suggests that the fitted model does not fully capture all influential variables or interactions within the system. Additionally, it was found that loading TNX into γ-CD-MOFs generally increased its water solubility, but small crystals showed faster drug dissolution, which could be attributed to their larger surface area compared to large crystals. This research not only enhances the understanding of γ-CD-MOFs as drug carriers but also sets the groundwork for further research aimed at enhancing the delivery of poorly water-soluble pharmaceuticals, paving the way for improved therapeutic strategies in clinical applications.